A Frequency-Domain Opportunistic Approach for Spectral-Efficient Cell-Free Massive MIMO

TL;DR

This work addresses the cell-edge QoS degradation in wideband cellular networks by proposing a frequency-domain opportunistic AP selection (OAS) for CFmMIMO-OFDM. By spreading users across OFDM RBs and selectively activating nearby APs while deactivating distant ones, the approach converts high-dimensional CFmMIMO into a set of low-dimensional, efficiently managed links and enables downlink pilots for coherent detection. The authors derive closed-form SE expressions for SU-OAS and MU-OAS and compare them against CF and UC baselines, demonstrating substantial uplink and downlink SE gains, particularly in the 5th percentile rate, with modest complexity and reduced fronthaul demands. This method offers significant practical benefits for spectral efficiency, power efficiency, and pilot overhead in future wideband CFmMIMO deployments.

Abstract

Constrained by weak signal strength and significant inter-cell interference, users located at the cell edge in a cellular network suffer from inferior service quality. Recently, cell-free massive MIMO (CFmMIMO) has gained considerable attention due to its capability to offer uniform quality of service, alleviating the cell-edge problem. In contrast to previous studies focused on narrow-band CFmMIMO systems, this paper studies wideband CFmMIMO communications against channel frequency selectivity. By exploiting the frequency-domain flexibility offered by orthogonal frequency-division multiplexing (OFDM), and leveraging a particular spatial characteristic in the cell-free structure -- namely, the near-far effect among distributed access points (APs) -- we propose an opportunistic approach to boost spectral efficiency. The core concept lies in opportunistically activating nearby APs for certain users across their assigned OFDM subcarriers while deactivating distant APs to prevent power wastage and lower inter-user interference. Furthermore, this approach enables the use of downlink pilots by reducing the number of active APs per subcarrier to a small subset, thereby substantially improving downlink performance through coherent detection at the user receiver. Verified by numerical results, our proposed approach demonstrates considerable performance improvement compared to the two benchmark approaches.

Figures (6)

• Figure 1: System model of a CFmMIMO-OFDM system that consists of a CPU, $M$ APs, and $K$ UEs. It illustrates the block diagram of end-to-end OFDM transmission between AP $m$ and UE $k$, where the DFT demodulator and IDFT modulator are implemented by fast Fourier transform (FFT) and inverse FFT (IFFT), respectively, when $N$ is a power of two.
• Figure 2: The time-frequency resource grid of a CFmMIMO-OFDM system. This figure shows an example where each resource block covers $N_{rb}=8$ subcarriers and spans the entire duration of a radio frame with $T=24$ OFDM symbols. Channel coefficients across all RUs can be acquired through channel estimation and interpolation with the lattice-type pilot pattern.
• Figure 3: Illustration of the near-far effect in a cellular system where a near user gets a strong signal from a BS whereas a far user suffers from a weak signal (left). In comparison, the distributed APs in a cell-free system offer strong and weak signals based on their distances to a user (right). They are labeled as a near AP or a far AP from the perspective of a specific user.
• Figure 4: Illustration of SU-OAS in a CFmMIMO-OFDM system, along with explanations of the pilot arrangement. To simplify comprehension, we present an example involving two users, each choosing four nearby APs.
• Figure 5: Performance comparison of the four approaches in a CFmMIMO-OFDM system, where a total of $256$ antennas are randomly distributed within a circular area with a radius of 1km to serve $16$ users.

Theorems & Definitions (4)

• Remark 1
• Theorem 1
• Theorem 2
• Remark 2